[news.eclipse.tools.hyades] test creating architecture

[Jan Topinski <simcha@xxxxxxxxxx>]

Hi,
I'm working in an internal project (in polish consulting firm Infovide) 
providing us whith testing framework. The project is run by QA team 
(part time). The project is in a very early stage. Week ago we 
discovered Hyades :), we liked it and decided to follow you - now we 
must decide what it means. For easer reading let us call our project 
Platforma(framework in polish).
I compared our architecture with Hyades. Platforma is of course much 
simpler but in general compatible. There is one point in which they seem 
a little bit diffirent - test creation (and execution in some way). 
Hyades  treats test creation as a whole. Creating (recording, 
generation) of interface model (I like the idea from thread "Autoated 
GUI testing") and   creating test model. And in case of Platforma we 
have interface model, test model(test scenario) and data model (nothing 
new yet). We planed to manage data in standarized form (DOM objects) so 
we can provide standard data compariton/syntax testing/generation 
functions, reuse data generators, and easer reuse of whole tests in 
diffirent AUT.
But the key diffirence is data generation module (we have limited 
functionality but working version). First you give the definition of 
"properly built object" (xml schema with extentions), so generator can 
randomly generate objects accepted(warning simplification!) by AUT. Then 
 you can define modifications to schema, so your objects fit specific 
business scenario needs or are not "properly built object" so you can 
test if AUT reject such object. This two (definition and modification) 
are separated to keep one current version of object definiton. When test 
is executed test engine can also make modification request to the 
generator (eg. to randomly regenerate part of object upon AUT answer).
Generation module is now in use in a project in which tests are disigned 
using methodology of data driven tests. Data driven tests are useful for 
eg. when testing batch processing, there is no (or very simple) test 
script totaly data dependent, all test logic is kept in data objects. 
Objects are randomly generated.
Hope this ideas will be useful. If you have some more answers on 
what/when questions it would be easer for us to decide how to "cope" 
with Hyades in our project (we will redesign it for sure).
Best regards,

                    Jan Topinski
                    (consultant)
                Infovide S.A. www.infovide.pl